lol (version 1.20.0)

lasso.multiSplit: Multi-split lasso

Description

Multi-split lasso as described in Meinshausen 2009

Usage

lasso.multiSplit(y, x=NULL, lambda1=NULL, nSubsampling=200, model='linear', alpha=0.05, gamma.min=0.05, gamma.max=0.95, track=FALSE, ...)

Arguments

y

A vector of gene expression of a probe, or a list object if x is NULL. In the latter case y should a list of two components y and x, y is a vector of expression and x is a matrix containing copy number variables

x

Either a matrix containing CN variables or NULL

nSubsampling

number of splits, default to 200

model

which model to use, one of "cox", "logistic", "linear", or "poisson". Default to 'linear'

alpha

specify significant level to determine the non-zero coefficients in the range of 0 and 1, default to 0.05

gamma.min

the lower bound of gamma

gamma.max

the higher bound of gamma

lambda1

minimum lambda to be used, if known

track

track progress

...

other parameters to be passed to lass.cv

Value

A list object of class 'lol', consisting of:

beta

coefficients

mat

the Q_gamma matrix as described in the paper

residuals

residuals, here is only the input y

pmat

the adjusted p matrix as described in the paper

Details

This function performs the multi-split lasso as proposed by Meinshausen et al. 2009. The samples are first randomly split into two disjoint sets, one of which is used to find non-zero coefficients with a regular lasso regression, then these non-zero coefficients are fitted to another sample set with OLS. The resulting p-values after multiple runs can then be aggregated using quantiles.

References

Nicolai Meinshausen, Lukas Meier and Peter Buehlmann (2009), P-values for high-dimensional regression. Journal of the American Statistical Association, 104, 1671-1681.

See Also

lasso

Examples

data(chin07)
data <- list(y=chin07$ge[1,], x=t(chin07$cn))
res <- lasso.multiSplit(data, nSubsampling=50)
res